Converting genetic activity into music may be a way to monitor health.
When set to music, colon cancer sounds kind of eerie. That’s the finding of Gil Alterovitz, a research fellow at Harvard Medical School who is developing a computer program that translates protein and gene expression into music. In his acoustic translation, harmony represents good health, and discord indicates disease.
At any given time in each of our cells, thousands of genes are churning out their molecular products while thousands more lie senescent. The profile of which genes are on versus off is constantly changing–with specific diseases such as cancer, for example.
Searching for a more simplified way to represent the complex library of information inherent in gene expression, Alterovitz decided to represent those changes with music. He hopes that doctors will one day be able to use his music to detect health-related changes in gene expression early via a musical slip into discord, potentially improving a patient’s outcome.
The first step in the gene-to-sound conversion was to pare down multiple measurements to a few fundamental signals, each of which could be represented by a different note. Together, the notes would form a harmonic chord in normal, healthy states and become increasingly out of tune as key physiological signs go awry, signaling disease.
Alterovitz employed mathematical modeling to determine relationships between physiological signals. Much like the various systems in an automobile, many physiological signs work in synchrony to keep a body healthy. “These signals [are] not isolated parts,” says Alterovitz. “Like in a car, one gear is working with other gears to control, for example, power steering. Similarly, there are lots of correlations between physiological variables. If heart rate is higher, other variables will move together in response, and you can simplify that redundancy and information.”
Using data collected from a study of protein expression in colon cancer, Alterovitz analyzed more than three thousand related proteins involved in the disease. He whittled down the thousands of proteins to four key networks, using various genetic databases that catalog relationships between genes and proteins. He then assigned a note to each network, and together, these notes formed a harmonic chord. He compared the “music” of normal, healthy human data sets to that of the colon-cancer samples and found that, according to his model, colon cancer sounded “inharmonious.”
Researchers may be able to translate other diseases into music by “tuning” the system that Alterovitz has developed. For example, researchers can identify protein networks related to the disorder of interest and then assign notes that, in combination, form inharmonious chords, compared with their healthy counterparts.
He adds that the technique may have applications outside medicine, such as for simplifying information for air-traffic controllers, and in any other industry that requires analysis of large data sets. There is also an opportunity to use protein music purely for music’s sake: a DJ in the Boston area has expressed interest in playing Alterovitz’s “music” in local bars–a potential revenue stream for musician and mathematician alike.